Show Notes: Why Algorithms Make Smart People Act Stupid

The Deep Dig on A Closer Look

Episode Description

In this episode of The Deep Dig, we break down Khayyam's incendiary op-ed The Stupidity Subsidy — a piece that doesn't just push back on the cultural panic around falling IQs, it flips the entire board over. The hosts begin where the doomers do: the data showing measurable IQ score declines across Norway, Denmark, France, and Britain. But rather than accepting the fashionable narrative that we are biologically regressing into stupidity, the episode methodically dismantles it, beginning with the landmark 2018 Bratsberg and Rogeberg sibling study — arguably the most elegant demolition of the genetic-rot theory in modern cognitive science.

From there, the episode pivots to economist Carlo Cipolla's long-forgotten 1976 pamphlet, The Basic Laws of Human Stupidity, and its radical proposition: stupidity has nothing to do with IQ. It is a coordinate on a payoff graph. Using Cipolla's four-quadrant behavioral map — intelligent (win-win), helpless (lose-win), bandit (win-lose), and stupid (lose-lose) — the hosts then walk through the mechanics of how modern algorithmic recommendation engines were engineered, whether by design or by default, to manufacture mass residency in that toxic bottom-left quadrant.

The episode goes further still, drawing on the Rathje PNAS study, the Duke University polarization experiment, and the ecological modeling of Parisi and Bardi to argue that the feed is not a rational bandit sustainably extracting from us — it is a parasitoid strip-mining the foundational lichen of shared human trust faster than it can regenerate. The central, chilling thesis: we didn't build a machine that makes us dumb. We built one that makes our intelligence irrelevant to our behavior — and then subsidized the gap at planetary scale.

Category / Topics / Subjects

• The Flynn Effect and Its Reversal
• Cognitive Science and IQ Research
• Cipolla's Economic Theory of Stupidity
• Behavioral Economics and Payoff Theory
• AI Alignment and Misaligned Reward Functions
• Social Media Algorithmic Design
• Outgroup Animosity and Viral Dynamics
• Ecological Modeling of Social Systems
• Moore's Law and the Saturation of Human-Machine Cognition
• Political Polarization and Filter Bubble Research
• The Attention Economy
• Civilizational Risk and Social Trust

Best Quotes

"The core argument here isn't that our inherent biological intelligence moved. It's that the reward function moved."

"We built a planetary-scale, trillion-dollar machine that literally pays brilliant people to act like absolute idiots. And then we blame the idiots for acting exactly how the machine paid them to act."

"Stupidity in this framework is a coordinate on a payoff graph, not a brainpower metric."

"The smarter you are, the better you are at being stupid."

"The machine structurally cannot optimize for calm because calm means you put the phone down."

"You cannot reason your way out from inside a misaligned field. Deliberation inside the machine doesn't make you empathetic. It just optimizes you deeper into defensive tribalism."

"We are stripping the lichen of shared reality down to the bedrock."

"The burning question was never whether we are biologically getting dumber. We aren't. The real question is why we built the first ubiquitous environment in human history, explicitly engineered to make our inherent intelligence entirely irrelevant to our behavioral outcomes — and then handed one to every single human alive. For free. Forever."

"You cannot untangle the Christmas lights if the machine is designed to keep tying the knots faster than your hands can work."

Three Major Areas of Critical Thinking

1. The Misdiagnosis Problem: Why the IQ Panic Is Wrong in the Right Direction

The episode asks a foundational epistemological question: when data shows declining cognitive test scores across multiple countries, what is the appropriate inference? The standard cultural narrative — we are biologically getting dumber — is emotionally satisfying and structurally lazy. The Bratsberg-Rogeberg sibling methodology is critical here precisely because it eliminates the standard confounders in one stroke: if the cognitive decline appears between brothers raised in the same house, the cause cannot be genetics, immigration composition, or differential parenting. It has to be environmental. The Dworak 2023 U.S. data adds a further complication: the picture isn't a uniform decline but a scramble — verbal reasoning and abstract logic slip while spatial reasoning rises. As the hosts note, this is adaptation, not degradation. The Flynn Paradox itself, where Flynn warned in 1987 that his own rising-scores data couldn't possibly reflect raw biological intelligence, provides the crucial theoretical anchor: IQ tests measure culturally contingent cognitive reflexes, not fixed brainpower. The richer critical question the episode invites is whether a society that has outsourced specific cognitive tasks to technology is losing a capacity or simply ceasing to practice a particular kind of attention — and what the normative implications of that distinction are.

2. The Architecture of Misalignment: How the Feed Was Engineered to Produce Stupidity

Cipolla's four-quadrant framework provides the episode's sharpest analytical edge. By defining stupidity as a behavioral outcome rather than a cognitive trait — actions that harm others while producing a net loss for the actor — Cipolla detaches the concept entirely from intelligence. The episode's central argument is that algorithmic recommendation engines don't just tolerate stupidity-quadrant behavior; they economically subsidize it. The Rathje PNAS study's finding that outgroup references increase share probability by 67% per word, and are 6.7 times more predictive of virality than moral-emotional language, makes the subsidy mechanism empirically explicit. This is the AI alignment problem made concrete and present: a system optimized for the proxy goal of engagement rather than the true goal of human flourishing will inevitably learn to manufacture outgroup contempt at industrial scale. The Duke University experiment — where forcing partisan users to follow opposing accounts increased polarization — closes the obvious escape hatch: you cannot cognitively override a system from within it. The episode's most unsettling implication is reserved for the intelligent: high verbal and analytical capacity, in a misaligned system, is a force multiplier for damage rather than a corrective. The smartest people are the most effective instruments of destruction precisely because their skill translates the machine's incentive into more articulate, weaponized, highly viral outrage.

3. The Ecological Ultimatum: What Happens When the Stupid Quadrant Becomes the Default Infrastructure

The episode's final and most sobering move is to ask what historical and biological modeling tells us about the long-term viability of a system that has enrolled its entire population in lose-lose dynamics. The reindeer of St. Matthew Island — 29 animals introduced in 1944, 6,000 by 1963, 42 skeletal survivors by 1966 — is deployed as more than metaphor; it is the literal output of the Lotka-Volterra equations applied to Cipolla's quadrant structure. The critical framework Parisi and Bardi propose is that the stupid quadrant in ecological terms is not predator-prey (which reaches sustainable equilibrium) but the overextracting parasite that kills its host and dies with it. The application to the attention economy is precise: human trust, shared epistemic reality, and civic coherence are the slow-growing lichen. The feed extracts quarterly engagement metrics by strip-mining that lichen at a rate no regeneration cycle can match — and unlike a fox, it has no line of code governing sustainable harvest. The episode identifies four structural features that make the current trap categorically different from historical precedents like tulip mania or the 2008 mortgage collapse: it is universal (4 billion users), continuous (every waking minute), personalized (real-time psychological profiling), and self-optimizing (machine learning tightens the grip without human direction). The final open question — whether we can engineer a profitable alignment algorithm, or whether evolutionary psychology makes outgroup contempt the inevitable product of any ad-revenue model — is deliberately left unresolved, functioning as the episode's true intellectual provocation rather than its conclusion.

For A Closer Look, click the link for our weekly collection.

::. \ W24 •A• The Stupidity Subsidy ✨ /.::

https://tokenwisdom-and-notebooklm.captivate.fm/episode/w24-a-the-stupidity-subsidy-

✨Copyright 2025 Token Wisdom ✨